Active Contour Models : William and Shah SnakeClive Saha, John Hsu and Chivas Nambiar |
|
|
Conclusions:In conclusion, we can say that while snakes have been shown to work on different types of images, the fact still remains that they are very sensitive to a large number of factors. The most important of these are the location and size of the initial snake structure. Automatic placement of the snake guarantees less than optimal performance. However, placing the starting points manually, though more accurate, is highly tedious. A number of preprocessing steps can be used to influence to placement of the snake initially, including use of different edge operators. But there are certain fields where snakes can be tweaked to find regions of particular interest in a class of images. This is especially true of medical images, eg: CT scan images. Also, while thresholding and region growing give us a simple method to achieve segmentation, snakes perform well under noisy conditions and with gradient background images. Back to Top |
|